Motion-Excited Sampler: Video Adversarial Attack with Sparked Prior

Zhang, H; Zhu, L; Zhu, Y; Yang, Y

Motion-Excited Sampler: Video Adversarial Attack with Sparked Prior

Zhang, H Zhu, L

Zhu, Y Yang, Y

Permalink

Publisher:: Springer
Publication Type:: Conference Proceeding
Citation:: Computer Vision – ECCV 2020 16th European Conference, 2020, 12365 LNCS, pp. 240-256
Issue Date:: 2020-01-01

Recently Added

	Filename	Description	Size
	Binder2.pdf	Accepted Manuscript	7.63 MB	Adobe PDF	View/Open

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is new to OPUS and is not currently available.

Full metadata record

Field	Value	Language
dc.contributor.author	Zhang, H
dc.contributor.author	Zhu, L https://orcid.org/0000-0002-4093-7557
dc.contributor.author	Zhu, Y
dc.contributor.author	Yang, Y
dc.date	2020-08-23
dc.date.accessioned	2021-04-21T06:45:52Z
dc.date.available	2021-04-21T06:45:52Z
dc.date.issued	2020-01-01
dc.identifier.citation	Computer Vision – ECCV 2020 16th European Conference, 2020, 12365 LNCS, pp. 240-256
dc.identifier.isbn	9783030585648
dc.identifier.issn	0302-9743
dc.identifier.issn	1611-3349
dc.identifier.uri	http://hdl.handle.net/10453/148252
dc.description.abstract	© 2020, Springer Nature Switzerland AG. Deep neural networks are known to be susceptible to adversarial noise, which is tiny and imperceptible perturbation. Most of previous works on adversarial attack mainly focus on image models, while the vulnerability of video models is less explored. In this paper, we aim to attack video models by utilizing intrinsic movement pattern and regional relative motion among video frames. We propose an effective motion-excited sampler to obtain motion-aware noise prior, which we term as sparked prior. Our sparked prior underlines frame correlations and utilizes video dynamics via relative motion. By using the sparked prior in gradient estimation, we can successfully attack a variety of video classification models with fewer number of queries. Extensive experimental results on four benchmark datasets validate the efficacy of our proposed method.
dc.language	en
dc.publisher	Springer
dc.relation	http://purl.org/au-research/grants/arc/DP200100938
dc.relation.ispartof	Computer Vision – ECCV 2020 16th European Conference
dc.relation.ispartof	European Conference on Computer Vision
dc.relation.ispartofseries	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
dc.relation.isbasedon	10.1007/978-3-030-58565-5_15
dc.rights	This is a post-peer-review, pre-copyedit version of an article published in [Computer Vision ECCV 2020 16th European Conference, 2020, 12365 LNCS, pp. 240-256]. The final authenticated version is available online at: https://link.springer.com/chapter/10.1007%2F978-3-030-58565-5_15/[insert DOI]”
dc.rights	info:eu-repo/semantics/restrictedAccess
dc.subject.classification	Artificial Intelligence & Image Processing
dc.title	Motion-Excited Sampler: Video Adversarial Attack with Sparked Prior
dc.type	Conference Proceeding
utslib.citation.volume	12365 LNCS
utslib.location.activity	Glasgow, UK
pubs.organisational-group	/University of Technology Sydney
pubs.organisational-group	/University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	/University of Technology Sydney/Strength - AAII - Australian Artificial Intelligence Institute
pubs.organisational-group	/University of Technology Sydney/Faculty of Engineering and Information Technology/School of Computer Science
utslib.copyright.status	recently_added	*
pubs.consider-herdc	false
dc.date.updated	2021-04-21T06:45:24Z
pubs.finish-date	2020-08-28
pubs.place-of-publication	Switzerland
pubs.publication-status	Published
pubs.start-date	2020-08-23
pubs.volume	12365 LNCS
dc.location	Switzerland

Abstract:

© 2020, Springer Nature Switzerland AG. Deep neural networks are known to be susceptible to adversarial noise, which is tiny and imperceptible perturbation. Most of previous works on adversarial attack mainly focus on image models, while the vulnerability of video models is less explored. In this paper, we aim to attack video models by utilizing intrinsic movement pattern and regional relative motion among video frames. We propose an effective motion-excited sampler to obtain motion-aware noise prior, which we term as sparked prior. Our sparked prior underlines frame correlations and utilizes video dynamics via relative motion. By using the sparked prior in gradient estimation, we can successfully attack a variety of video classification models with fewer number of queries. Extensive experimental results on four benchmark datasets validate the efficacy of our proposed method.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/148252